Training regimens, manifolds, and vector fields

Resumen

A main question for sport scientists is: How can training methods for perceptual and perceptual-motor skills be optimized? This presentation summarizes recent research on learning that suggests the following methodology. One first aims to identify a low-dimensional manifold that describes different ways in which the task can be performed. To give an example: if one studies perceptual learning each point in the manifold may represent an informational variable. Learners that use a particular informational variable can then be localized at a point in the manifold, and a group of learners, at a particular phase of the training, can be localized by a probability distribution on the manifold. Change that occurs during learning for a group of learners is hence described as a change in the probability distribution: After practice one expects more learners to be localized at the better loci. This change in the probability distribution, in turn, can be described by a vector field on the manifold. Our methodology allows us to calculate the practice conditions that optimize this vector field, and hence optimize the predicted rate of learning.